HAL Id: hal-00504077
https://hal.archives-ouvertes.fr/hal-00504077
Submitted on 19 Jul 2010
HAL is a multi-disciplinary open access archive for the deposit and dissemination of sci- entific research documents, whether they are pub- lished or not. The documents may come from teaching and research institutions in France or abroad, or from public or private research centers.
L’archive ouverte pluridisciplinaire HAL, est destinée au dépôt et à la diffusion de documents scientifiques de niveau recherche, publiés ou non, émanant des établissements d’enseignement et de recherche français ou étrangers, des laboratoires publics ou privés.
beech forests
Arnault Lalanne, Jacques Bardat, Fouzia Lalanne-Amara, Jean-François Ponge
To cite this version:
Arnault Lalanne, Jacques Bardat, Fouzia Lalanne-Amara, Jean-François Ponge. Local and re- gional trends in the ground vegetation of beech forests. Flora, Elsevier, 2010, 205 (7), pp.484-498.
�10.1016/j.flora.2009.12.032�. �hal-00504077�
Local and regional trends in the ground vegetation of beech forests
1 2
Arnault Lalanne1, Jacques Bardat2, Fouzia Lalanne-Amara1, Jean-François Ponge1* 3
4
1Muséum National d’Histoire Naturelle, CNRS UMR 7179, 4 avenue du Petit-Château, 91800 5
Brunoy, France; e-mail ponge@mnhn 6
2Muséum National d’Histoire Naturelle, CNRS UMR 7205, 57 rue Cuvier, Case Postale 39, 7
75231 Paris Cédex 05, France 8
9
*Corresponding author 10
11
E-mail addresses of the authors:
12 13
A. Lalanne, F. Lalanne-Amara: lalanne@mnhn.fr 14
J. Bardat: bardat@mnhn.fr 15
J.F. Ponge: ponge@mnhn.fr 16
17
Abstract
1 2
We sampled moss and vascular forest vegetation in five ancient beech forests from 3
northwest France, embracing in each a wide array of environmental conditions. Indirect 4
(PCA) and direct (RDA) gradient analysis were used to discern local and regional ecological 5
factors which explain the observed variation in species composition. Our results point to a 6
global factor encompassing a large array of soil and light conditions, unravelled when local 7
particularities of studied forests are partialled out. The humus form, numerically expressed by 8
the Humus Index, explains a large part of the observed variation in ground vegetation. Our 9
study confirmed opposite trends in vascular and moss species richness according to humus 10
condition. Ecological factors to which vascular and moss forest species respond at the 11
regional level can be estimated directly on the field by visually inspecting humus forms and 12
vegetation strata despite of the confounding influence of local factors.
13 14
Keywords: Beech forests; Ground vegetation; Humus Index; Regional and local factors;
15
Species groups; Species richness 16
17
Introduction
18 19
It is widely recognized that forest vegetation varies according to soil fertility and 20
acidity, temperature, light and moisture (Watt, 1934; Ellenberg, 1974; Diekmann and 21
Lawesson, 1999), and is a fairly good indicator of potential growth (Bergès et al. 2006) and 22
ecological integrity (Aubin et al., 2007) of managed forests. It is also admitted that 23
management practices may modify the composition of understory plant communities through 24
their influence on the abovementioned factors (Thomas et al., 1999; Gillet et al., 1999; Van 25
Calster et al., 2007). Once the ecological requirements of plant species are documented and 1
indexed over wide environmental gradients (Diekmann, 2003), then lists of plant species and 2
their ecograms (Härdtle et al., 2004) can be used to achieve a thorough assessment of site 3
conditions, often in conjunction with humus types (Bartoli et al., 2000; Wilson et al., 2001, 4
but see Wamelink et al., 2002). However, the composition of forest plant communities may 5
also vary locally due to historical heritage (Dupouey et al., 2002; Gachet et al., 2007), 6
dispersal limitation (Bossuyt et al., 1999; Graae and Sunde, 2000; Honnay et al., 2002), biotic 7
interactions (Thompson et al., 1993; Britton et al., 2003; Klanderud and Totland, 2007) and 8
landscape features (Dufour et al., 2006), and thus is in a constant state of change (Wiens, 9
1984). Among a regional pool, each forest, for past as well as present-day reasons, will select 10
or favour a subsample of species and species traits which will be redistributed according to 11
niche requirements of species, assembly rules and disturbance effects (Keddy, 1992; Zobel, 12
1997; Rajaniemi et al., 2006). Stable although distinct forest communities may develop on 13
similar sites (McCune and Allen, 1985) as a consequence of attraction domains (Beisner et al., 14
2003).
15 16
We need tools for disentangling the joint influence of stable and variable 17
environmental factors on vegetation (Ehrenfeld et al., 1997) and to discern environmental 18
gradients of regional importance when plant communities vary locally to a great extent 19
(Ricklefs, 1987; Huston, 1999; Hillebrand, 2005): does the presence of a species indicate the 20
same thing in one and another forest? The present study is an attempt to fulfil this gap within 21
the domain of European beech (Fagus sylvatica L.) forests. For that purpose we selected five 22
ancient forests (Peterken and Game, 1984; Hermy et al., 1999) in Northwest France, where 23
two Atlantic forest habitats, the neutrophile Endymio-Fagetum or EU priority habitat type 24
41.132 (EUR25, 2003) and the acidiphilous Ilici-Fagetum or EU priority habitat type 41.12 25
(EUR25, 2003), are well-represented (Bardat, 1993). These beech habitats were described in 1
West and Northwest France as two associations by Durin et al. (1967). We sampled moss and 2
vascular forest vegetation in 995 plots (400 m2) embracing a wide array of canopy, 3
understory, stand age, humus type, soil disturbance, climate and atmospheric deposition 4
conditions. Indirect (PCA) and direct (RDA) gradient analysis were used to discern main local 5
and regional ecological factors which could explain the observed variation in species 6
composition. A particular attention will be given to the humus type, which is both a cause and 7
a consequence of vegetation development (Ponge, 2003; Godefroid et al., 2005) and to which 8
beech forest plant species are highly sensitive (Le Tacon and Timbal, 1973; Falkengren- 9
Grerup and Tyler, 1993; Lalanne et al., 2008).
10 11
Materials and methods
12 13
Study sites 14
15
Five forests, attested at least from the Roman period (except Compiègne attested from 16
the 14th century only) have been selected in Picardy and Upper Normandy (Table 1). The 17
choice of these forests was dictated by the need to assess the influence of local particularities 18
in the studied region, while remaining in the same plant associations, Endymio-Fagetum (EF) 19
and Ilici-Fagetum (IF) on neutral to acid soils, respectively. All these forests belonged to the 20
royal domain then acquired a national status after the French revolution. After the strong 21
deforestation which occurred during Middle Ages they were submitted to the Forest Law and 22
were managed as coppices-with-standards from 16th to 19th century, then they were 23
progressively converted to full-grown mixed forests. European beech is dominant, but sessile 24
oak (Quercus petraea) or pedunculate oak (Quercus robur) are subordinate or co-dominant in 25
the canopy. All these lowland forests (altitude < 250 m) are established on cretaceous 1
limestone tables of the ‘Bassin Parisien’. The calcareous substrate is covered with tertiary and 2
quaternary deposits of varying depth and nature, which were eroded on slopes, providing a 3
variety of strongly acid to alkaline soils. The Brotonne forest is established mainly on a fossil 4
meander of river Seine made up of many quaternary gravel-sand alluvial terraces dated from 5
Riss to Wurm (Quaternary Age). Eawy and Lyons forests occupy especially plates with a 6
thick cover of loess (Quaternary Age). The Compiegne forest is located on a sandy slope 7
(Cuisian stage) and on Sparnacian marls of the alluvial river Oise valley (Tertiary Age). The 8
Retz forest is established on loess (Quaternary Age) and sand (Tertiary Age, Stampian stage) 9
deposits. The annual rainfall decreases and the mean temperature increases from Brotonne 10
and Eawy to Compiègne and Retz, Lyons being intermediate, according to a decreasing 11
Atlantic influence from West to East, without any marked North-South trend (Table 1).
12
Sulphur deposition is higher in Brotonne and Eawy, which are not far from oil refineries 13
located in Le Havre and Rouen, respectively. Nitrogen deposition, mostly of industrial origin 14
in Eawy, and of agricultural origin in Retz, is higher in these two forests. Values reported on 15
Table 1 were interpolated from a national grid of continuous measurements of atmospheric 16
dry deposition, except for Brotonne where measurement was direct (Croisé et al., 2005).
17 18
As abovementioned, our study focused on two types of beech habitats, in which beech 19
is associated with holly (Ilex aquifolium) in the understory on most acidic soils (upslope sites 20
with Podzols and Luvisols at pH < 5) and with bluebell (Hyacinthoides non-scripta) on 21
moderately acidic to neutral soils (downslope sites with Luvisols and Cambisols at pH 22
between 5 and 7). Both habitats are present in the five studied forests, but their species 23
composition varies locally according to climate and geomorphology. Bardat (1993) described 24
numerous sub-associations and variants (‘sylvofacies’) on the basis of moss and vascular 25
forest vegetation. Following preliminary investigations, study sites were selected in each 1
forest in order to encompass the widest possible variety of environmental, management and 2
stand age conditions. More sites were sampled in Eawy, the smallest forest investigated, in 3
order to compensate for the relative homogeneity of site conditions. The choice of a non- 4
random selection of forests and stands within forests was dictated by the need to avoid severe 5
biases in the representativeness of sampling sites. More details about the selection of sites 6
according to stand age classes are given in Lalanne (2006).
7 8
Sampling design and data analysis 9
10
At each site five plots 20 x 20 m, four at angles of a 1 ha square and one at its centre 11
(Lalanne et al. 2008) were surveyed for ground vegetation and environmental factors. In each 12
plot vascular plants (herbs, ferns) and mosses were identified at the species level (Appendix 13
1) and quantified according to the Braun-Blanquet method. Kerguelen (1993), lastly updated 14
at http://www2.dijon.inra.fr/flore-france/index.htm, was used for the nomenclature of vascular 15
plants, and Hill et al. (2006) for mosses. Species living in aboveground micro-habitats (trunk 16
bases, boulders, dead branches and trunks) were not recorded. Previous to numerical 17
treatment Braun-Blanquet cover-abundance data were transformed in percentage values 18
according to Van der Maarel (1979). Species which were present in less than 10 sample plots 19
were excluded from further analyses, except for the calculation of species richness.
20 21
Ellenberg’s indices (Light, Wetness, Fertility, pH) were affected to each plant species 22
(Appendix 1). For vascular plant species we used indicator values calculated for the British 23
Isles (Hill et al., 1999). For moss species we used a table prepared by one of us (J.B.), which 24
was partly published in Bardat and Aubert (2007). The Fertility Index was equivalent to the N 25
(‘Stickstoff’) index of Ellenberg (1974). Ellenberg’s indices were averaged for each plot 1
according to Wamelink and Van Dobben (2003), in order to provide a global assessment of 2
ecological requirements of vegetation units on the base of auto-ecological characters.
3 4
Environmental descriptors were recorded at each sampling plot (Appendix 2). Canopy 5
cover, as well as sub-canopy cover and shrub cover when present, were estimated by cover- 6
abundance values according to the Braun-Blanquet scale and subsequently transformed into 7
cover percentages as abovementioned. The Humus Index was estimated by scoring humus 8
forms according to Ponge et al. (2002). Four measurements of the Humus Index were 9
averaged after subdividing each 400 m2 plot into four 100 m2 sub-plots, at the centre of which 10
as small pit was dug off. The choice of a regular grid rather than of a random selection of 11
points was due to the combined need for representativeness and minimum digging of the soil.
12
The ground surface was thoroughly observed and classified into several categories of topsoil 13
disturbance and dead wood deposition (Appendix 2), the cover percentage of which was 14
estimated visually in each plot. For the need of calculation, categories of topsoil disturbance 15
were pooled into gross categories (undisturbed = Surf1, weak disturbance = Surf2 to Surf8, 16
severe disturbance = Surf9 to Surf14). Stem density and basal area were estimated by 17
counting and measuring the diameter at breast height (DBH) of all trees and shrubs with a 18
stem diameter higher than 7.5 cm. The age of dominant trees (crowns extending above the 19
general level of the canopy) was estimated by classifying them into ‘regeneration’, 20-40 20
years, 70-90 years, 120-140 years and 170-200 years.
21 22
The whole ground vegetation data set was submitted to Principal Components 23
Analysis (PCA), after standardization of cover percentage values to mean = 0 and variance 24
=1, thereby equalling Euclidean distances between plant species to product-moment 25
correlation coefficients. Plant species, coded as in Appendix 1, were projected on bi-plots of 1
factorial axes. Other bi-plots were used to show the projection of some additional variables on 2
the same factorial axes. Although somewhat neglected by plant ecologists after the rise of 3
methods derived from Correspondence Analysis (CoA), PCA with standardized variables was 4
chosen because of the ease with which it may reveal gradients as well as clusters in a complex 5
data matrix, without being influenced by rare species nor by preconceived hypotheses (Chae 6
and Warde, 2006; Bakkestuen et al., 2008). Comparisons done by Kenkel (2006) over a large 7
array of multivariate methods, including the commonly used Detrended Correspondence 8
Analysis (DCA) concluded to the superiority of PCA to decrease the number of dimensions of 9
large data sets. Ellenberg’s indices and several measured environmental descriptors were 10
projected as additional (passive) variables in order to facilitate the ecological interpretation of 11
PCA axes.
12 13
In each studied forest and in each beech habitat (Endymio-Fagetum and Ilici-Fagetum) 14
the between-sample floristic variation was measured by summing up variance components of 15
sample scores along the first three PCA components. This was used as a measure of β- 16
diversity according to Ter Braak (1983).
17 18
In order to verify whether gradients and clusters revealed by indirect gradient analysis 19
(PCA) could be defined on the basis of environmental parameters measured at each sampling 20
plot a Redundancy Analysis (RDA) was performed, using two matrices, (i) ground vegetation 21
data used for PCA, (ii) environmental data listed in Appendix 2. Note that Ellenberg’s indices 22
were not used in this analysis, given that they give only an indirect view of the environment.
23
Thereafter a partial Redundancy Analysis (partial RDA) was performed, in order to verify that 24
the composition of ground vegetation could be explained by stand and ground properties 25
when the confounding influence of geographical distance was discarded. For that purpose the 1
five forests were added as five qualitative variables which were coded as 1 or 0 according to 2
sample location.
3 4
Given that spatial autocorrelation was expected due to our nested sampling design, we 5
used Signed Mantel tests (Oberrath and Böhning-Gaese, 2001) to investigate several plant- 6
environment relationships within each of the studied forests.
7 8
All calculations were done under Microsoft® EXCEL® using the Addinsoft®
9
XLSTAT® statistical software.
10 11
Results
12 13
Principal Components Analysis: the F1-F2 bi-plot 14
15
The vegetation data matrix which was analysed by PCA was comprised of 995 rows 16
(sample plots) and 141 columns (species). The first three components of PCA (those which 17
were interpretable in terms of ecological factors as a rule of thumb) extracted 11% of the total 18
variance, a weak although significant percentage (Kaiser criterium for eigen values and 19
Bartlett’s test of sphericity) which is explained by the high number of variables (141) 20
included in the analysis and the ground noise caused by scarcely represented plant species.
21
The projection of plant species in the plane of the first two components of PCA (eigen values 22
7.61 and 4.55, respectively) showed three directions over which the cloud of species was 23
stretched (here called ‘branches’), which were noted A, B and C (Fig. 1). The A branch was 24
stretched along the positive side of factor F1. Species projected far from the origin along the 25
A branch were considered characteristic of this branch and were noted as A group (Table 2).
1
All these species were positively correlated with F1. No branch was visible on the negative 2
side of this factor. Branches B and C were stretched on opposite sides of F2, not far from the 3
origin along F1. Two groups of characteristic species, which were positively and negatively 4
correlated with F2, respectively, were noted as B and C groups (Table 2). Fagus sylvatica 5
(Fga) and Rubus fruticosus agg. (Rfr) were projected in an intermediate position between A 6
and B branches, both species being correlated positively with F1.
7 8
The calculation of variable scores (Table 3) and the projection (Fig. 1) of additional 9
variables in the F1-F2 species plane showed that F1 was positively correlated with Fertility, 10
pH and Wetness Ellenberg’s indices, while it was negatively correlated with Ellenberg’s Light 11
index and Humus Index. Among canopy and sub-canopy descriptors, F1 was positively 12
correlated with canopy and sub-canopy hornbeam cover and total canopy cover. Factor F1 13
was positively correlated with vascular and total species richness and negatively with moss 14
richness. Gross disturbance categories (undisturbed, weak disturbance, severe disturbance) of 15
the ground floor were poorly correlated with F1.
16 17
Factor F2 was positively correlated with Humus Index and with Ellenberg’s Light 18
index (Table 3, Fig. 1), while it was negatively correlated with pH and Fertility Ellenberg’s 19
indices. Among canopy and sub-canopy descriptors, F1 was positively correlated with canopy 20
beech cover and total canopy cover. Factor F1 was negatively correlated with vascular and 21
total species richness. Gross disturbance categories were poorly correlated with F1.
22 23
The projection of samples in the F1-F2 bi-plot showed that A, B and C species groups 24
were not equally distributed in the five studied forests and in the two studied beech habitats 25
(Fig. 2, compare with Fig. 1). The B group (positive values of F2) was mostly represented in 1
the Ilici-Fagetum (IF) and its characteristic species were better displayed in Brotonne. The A 2
and C groups (positive values of F1 and negative values of F2, respectively) were mostly 3
represented in the Endymio-Fagetum (EF) and their characteristic species were better 4
displayed in Retz and Compiègne, respectively. Lyons EF was intermediate between A and C 5
species branches. It should be noted that EF and IF formed a continuum, without any clear-cut 6
limit between them, and that Brotonne EF was projected at the same position (i.e. exhibited 7
the same correlation with F1 and F2) than Eawy IF and Retz IF.
8 9
Principal Components Analysis: the F2-F3 bi-plot 10
11
While B and C branches were projected on opposite sides of F2, as mentioned above, 12
an additional branch D was displayed on the negative side of F3, while B and C were both 13
projected on its positive side (Fig. 3). The characteristic group of species corresponding to the 14
D branch (D group) shared two species with the A group: Lamium galeobdolon agg. (Lga) 15
and Athyrium filix-femina (Afi) (Table 2).
16 17
Contrary to F1 and F2, factor F3 (eigen value 3.42) was associated with ground 18
disturbance cover variables, being positively correlated with weak and severe disturbance 19
cover and negatively correlated with undisturbed cover (Table 3, Fig. 3). Factor F3 was 20
negatively correlated with moss species richness and subcanopy cover.
21 22
The projection of samples on the F2-F3 bi-plot (Fig. 4, compare with Fig. 3) showed 23
that the D group (negative values of F3) was mostly represented in Lyons EF and Eawy EF.
24
Other factors were not considered, because they did not exhibit any new species group, but 25
rather subdivided and reassembled A, B, C and D branches, without any clear-cut association 1
with environmental variables. We considered them as belonging to ground noise.
2 3
Floristic heterogeneity 4
5
Table 4 showed that floristic heterogeneity, as ascertained from the variance of sample 6
scores along the first three components of PCA, varied to a great extent according to forests 7
and according to EF and IF beech habitats. More between-sample variation was depicted by 8
the Endymio-Fagetum compared to the Ilici-Fagetum in Compiègne, Retz and Lyons. Eawy 9
remained at a very low level of floristic heterogeneity, whatever beech habitats and despite a 10
higher sampling effort (Table 1). In Brotonne the Ilici-Fagetum exhibited a higher floristic 11
heterogeneity than the Endymio-Fagetum.
12 13
Redundancy Analysis 14
15
Permutation tests (Monte-Carlo simulation) showed that the vegetation data matrix 16
and the environmental data matrix were not independent (Pseudo-F = 0.31, P < 0.0001) 17
despite of the fact that constrained variation (explained by environmental variables) was only 18
13% of total variation. The first three canonical axes extracted 46% of the constrained 19
variation (22%, 14%, 10%, respectively). Indirect and direct gradient analyses gave similar 20
results. A comparison between corresponding species bi-plots of RDA (Fig. 5) and PCA 21
(Figs. 1, 3) showed that the same four species groups were depicted by both analyses.
22
Correlation coefficients between RDA scores and PCA coordinates of plant species along the 23
first three axes were highly significant (r = 0.94, 0.93, 0.72, respectively, all with P < 0.0001).
24 25
A much simpler pattern was exhibited when the effect of geographical distance (and 1
associated factors such as climate and past history) was eliminated (partial RDA). The two 2
habitats EF and IF were clearly separated along Axis 1 (17.6% of constrained variation), 3
Endymio-Fagetum samples being projected on its positive side while Ilici-Fagetum samples 4
were projected on its negative side (Fig. 6). Among forest vegetation plant species, Milium 5
effusum (Mef), Athyrium filix-femina (Afi), Lamium galeobdolon agg. (Lga), Kindbergia 6
praelonga (Kpr), Hyacinthoides non-scriptus (Hno) and Carpinus betulus (Cpe) were 7
projected far from the origin on the positive side of Axis 1, while Ilex aquifolium (Iaq), 8
Pteridium aquilinum (Paq), Deschampsia flexuosa (Dfl), Molinia caerulea (Mca), Carex 9
pilulifera (Cpi) and Polytrichastrum formosum (Pfo) were projected far from the origin on its 10
negative side (Fig. 6). Mantel tests on the set of plant species showed that Ellenberg’s indices 11
for pH and Fertility were positively correlated with Axis 1 (rM = 0.41, P < 0.001 and rM = 12
0.31, P < 0.001, respectively), and Light Index was negatively correlated with the same axis 13
(rM = -0.31, P < 0.001). The Wetness index exhibited a weak (although significant) negative 14
correlation with Axis 1 (rM = -0.02, P < 0.01).
15 16
The projection of environmental variables (listed in Appendix 2) showed that Axis 1 17
of partial RDA was positively correlated with hornbeam canopy cover (Cov1, r = 0.36) and 18
hornbeam subcanopy cover (Cov5, r = 0.23), and was negatively correlated with Humus 19
Index (H.I., r = -0.42) and holly (Ilex aquifolium) shrub cover (Cov14, r = -0.31).
20 21
The architecture of the cloud of samples in the plane of the first two axes of partial 22
RDA (Fig. 6) showed a parabolic arrangement which is reminiscent of a Guttman effect (also 23
called ‘arch’ or ‘horseshoe’ effect), well-known in gradient analysis when most significant 24
information is provided by the first canonical axis (Hill and Gauch, 1980). As a consequence 1
further canonical axes were not taken into account: they were considered as ground noise.
2 3
The indicator value of the Humus Index at regional and local level 4
5
Among other environmental variables measured in our study sites, the Humus Index 6
was given a prominent position by partial RDA: its strong negative score along the main 7
gradient (Axis 1) and the subordinate gradient (Axis 2) depicted by this analysis indicated that 8
it explained a large part of variation in the composition of beech forest plant habitats. We 9
tested the relationships between the Humus Index and several community indices calculated 10
at the sample level: species richness of vascular plants and mosses, average Ellenberg’s 11
indices and percent cover by species belonging to the four groups depicted by PCA. In order 12
to avoid the confounding influence of discrepancies between forests (as exemplified by PCA 13
bi-plots) Mantel statistics were calculated in each forest, separately (Table 5). In all five 14
forests the Humus Index was positively correlated with moss richness and negatively 15
correlated with vascular richness. In Eawy the correlation was weak, although still significant.
16
In all forests, the Humus Index was negatively correlated with pH and Fertility Ellenberg’s 17
indices. The correlation between the Humus Index and the Light Ellenberg’s index was 18
positive in four out of five forests, Eawy being the exception with a weak (although 19
significant) negative value. The correlation between the Humus Index and the Wetness 20
Ellenberg’s index was negative in four out of five forests, Retz being the exception with a 21
weak (although significant) positive value).
22 23
The four groups of species depicted by PCA exhibited homogeneous trends whatever 24
the forest considered. The surface covered by species belonging to A, C and D groups was 25
negatively correlated with the Humus Index, while the correlation was positive for the B 1
group (Table 5).
2 3
Discussion
4 5
Our study confirmed that the humus form, considered as a synthetic indicator of the 6
soil nutrient regime (Wilson et al., 2001; Pyatt et al., 2001; Ponge et al., 2002), can explain a 7
large part of the observed variation in the floristic composition of beech forests established on 8
acid to neutral soils. Over a large set of beech stands, belonging to two common habitats, 9
Endymio-Fagetum (EF) and Ilici-Fagetum (IF), our study confirmed the opposite trends in 10
vascular and moss species richness which had been shown in Brotonne and Saint-Palais IF 11
beech stands (Lalanne et al., 2008): at the sampling plot level, when the Humus Index 12
increases, passing from mull (Humus Index 1-4) to moder (Humus Index 5-7) then to mor 13
(Humus Index 8), moss species richness increases and vascular species richness decreases, 14
thereby reinforcing the view of phylogenetic conservatism of the ecological niche (Prinzing et 15
al., 2001). This observation is in agreement with Brunet et al. (1997) and Roem and Berendse 16
(2000) who showed that at local level vascular plant species richness decreases when soil 17
acidity increases in woodland and grassland, respectively. It should be highlighted, however, 18
that this observation holds for plot scale only and does not preclude the existence of an 19
opposite trend at larger scales (Levin, 2000; Gering and Crist, 2002; Rajaniemi et al., 2006).
20
Given a regional species pool (Zobel, 1997) more fertile, less acidic soils, will allow more 21
vascular species to cohabit at small scale and share a dearth of nutrients through common 22
mycelial networks and facilitation mechanisms (Bruno et al., 2002; Hart et al., 2003).
23 24
In all studied forests; the Humus Index was positively correlated with pH and Fertility 1
Ellenberg’s indices, reinforcing the view expressed by Ulrich (1994) and Ponge (2003) that 2
humus forms, nutrient levels and soil acidity are strongly interconnected and interact with 3
vegetation types according to a limited number of ecosystem strategies (Odum, 1969) or 4
assembly rules (Belya and Lancaster, 1999). In a comprehensive study of understory 5
vegetation comparing 99 even-aged oak forests located in northern France, Bergès et al.
6
(2006) showed that humus forms, Ellenberg’s Reaction and Nitrogen indices (here called pH 7
and Fertility), H+, Ca and P concentrations explained the main part of the observed floristic 8
variation. Similar results were obtained by Wilson et al. (2001) in 70 sites covering the range 9
of soil nutrient conditions prevailing in plantation forests of the United Kingdom. In both 10
cases nitrogen acted as a pollutant rather than as a nutrient and was not associated with this 11
soil fertility gradient. This further justifies that the ‘Nitrogen index’ of Ellenberg (1974) was 12
renamed ‘Fertility Index’ by Hill et al. (1999).
13 14
In all but one studied forests the Humus Index was positively and negatively 15
correlated with Light and Wetness Ellenberg’s indices, respectively. In terms of causality this 16
seems to indicate that moisture disfavours and light favours the accumulation of organic 17
matter. That optimum (but not excess) moisture availability favours the circulation of 18
nutrients is a matter of fact (Austin et al., 2004) and the exception of the Retz forest, which 19
does not follow this general trend, can be explained by the high frequency of waterlogging, 20
which is currently associated with organic matter accumulation (Låg, 1971; Tate et al., 1995).
21
The observed (and at first sight surprising) correlation between light and organic matter 22
accumulation can be explained if we take into account that beech stands on more acidic soils 23
are less productive (Ponge et al., 1997), have less hornbeam and more oak, bracken and 24
heather in the understory (Lawesson, 2000), thereby shed more light on the ground. If this 25
hypothesis is true, it can be said that soil conditions (including humus form, acidity and 1
nutrient level) influence light conditions (Endler, 1993) by decreasing the influence of shade- 2
casting easy-to-decay species such as hornbeam and favouring light-demanding hard-to-decay 3
species such as oak, bracken, bilberry and heather, thereby reinforcing contrasts between 4
vegetation types (Miles, 1985). Rather than simple causal effects, our results point to the 5
existence of a global factor encompassing a large array of microclimate and soil conditions, 6
which will be discussed further below.
7 8
Among the four groups of forest vegetation species depicted by PCA in beech forests 9
of north-western France, the B group was the only one to be positively correlated with the 10
Humus Index, thus to be favoured by more acidic soils with a high Humus Index (mor and 11
moder). Species of this group (Carex pilulifera, Deschampsia flexuosa, Holcus mollis, Ilex 12
aquifolium, Lonicera periclymenum, Polytrichastrum formosum, Pteridium aquilinum, 13
Thuidium tamariscinum) are known for a long time for their preference for acid soils in a 14
wide array of oak and beech European forests (Olsen, 1925; Le Tacon and Timbal, 1973;
15
Ellenberg, 1974) and H-concentration was shown to be the driving factor (Falkengren-Grerup 16
and Tyler, 1993), although many other related processes may be involved, too (Lee 1999).
17
Among the sub-associations and variants of the Ilici-Fagetum described by Bardat (1993) in 18
the same region, this group is better represented by the Ilici-Fagetum holcetosum, which is 19
typically observed in forests on alluvial terraces of the Seine valley (Brotonne, Eawy, Lyons).
20
This is congruent with our results (Figs. 1, 2) except that this group was also present in Retz 21
IF and at least partially in Compiègne IF, two forests which were not covered by the study 22
cited above. Despite the vascular vs moss trend mentioned above, acidotolerant species of the 23
B group are not phylogenetically and physiognomically related (a shrub, a liana, two grasses, 24
a sedge-grass, a fern and two mosses). Within the limits of present knowledge, they rather 25
have in common to be perennial, to exhibit strong interference competition and anti-herbivore 1
defence (Jarvis, 1964; Coley et al., 1985; Dolling et al., 1994) and to excrete a variety of 2
organic acids with chelating properties in reaction to Al-toxicity (Schöttelndreier et al., 2001).
3
These characters stem in patch occupancy (Ovington, 1953; Watt, 1976), thereby decreasing 4
species coexistence at the plot level while allowing it over wide areas.
5 6
Beside these results, the case of Brotonne should be highlighted. Figure 2 showed that 7
most samples taken in the Endymio-Fagetum had positive values for F2, thus exhibited a 8
floristic composition which did not resemble that of the same habitat in other forests of the 9
same region. The shift of Brotonne EF towards an acidotolerant species distribution (B 10
branch) can be explained by the particular abundance of Deschampsia flexuosa, known to 11
thrive in the presence of atmospheric deposition of varied origin (Scale, 1980; Falkengren- 12
Grerup, 1986; Britton et al., 2003), in this air-polluted forest downwind of Le Havre (Solmon 13
et al., 2004; Croisé et al., 2005). The better representation of acidotolerant vegetation in 14
Brotonne was also reflected in a higher level of floristic variation (β-diversity) of the Ilici- 15
Fagetum (Table 4).
16 17
All other species groups exhibited by PCA (A, C, D) were negatively correlated with 18
the Humus Index, thus they were favoured by mull and less acidic soil conditions. However, 19
they diverged in their species composition, according to other environmental or geographical 20
factors which act as filters selecting subsets of species within regional pools (Weiher and 21
Keddy, 1995).
22 23
The A group (Table 2) included a liana (Hedera helix), three grasses (Melica uniflora, 24
Milium effusum, Poa nemoralis), three sedge-grasses (Carex pendula, C. remota, C.
25
sylvatica), ten forbs, among which two were annual (Geranium robertianum, Moehringia 1
trinervia) and eight were perennial (Circaea lutetiana, Euphorbia amygdaloides, Galium 2
odoratum, Geum urbanum, Lamium galeobdolon agg., Oxalis acetosella, Veronica montana, 3
Viola reichenbachiana) and two ferns (Athyrium filix-femina, Dryopteris filix-mas). All these 4
plants are commonly associated with nutrient-rich and moderately moist environments 5
(Rameau et al. 1989). This group is better represented by the Endymio-Fagetum typicum, 6
typically observed on loess deposits north of the river Seine (Bardat 1993). PCA showed that 7
it was mostly present in Retz EF (Fig. 2), which exhibited a high level of β-diversity (Table 8
4).
9 10
The C group included three trees (Acer campestre, A. platanoides, Tilia platyphyllos), 11
two shrubs (Crataegus monogyna, Evonymus europaeus), eight forbs, among which four were 12
annual or biennial (Aethusa cynapium, Alliaria petiolata, Galium aparine, Mycelis muralis) 13
and four were perennial (Fragaria vesca, Mercurialis perennis, Potentilla sterilis, Urtica 14
dioica), and a moss (Rhytidiadelphus squarrosus). These plants are commonly associated with 15
nutrient-rich soils, but they tolerate drier conditions than the A group and are often found on 16
calcareous soils (Rameau et al., 1989). The C group was better represented by the Mercurialo- 17
Aceretum (Bardat, 1993), which is a xerothermophilic and basocline association, located on 18
well-drained colluvial soils with admixture of limestone. It is known for its high vascular 19
richness, as attested by the projection of the corresponding vector in the F1-F2 plane (Fig. 1).
20
The C group was mostly present in Compiègne EF (Figures 3, 4).
21 22
The increase in the part played by annual herbs (0→2→4) along the B→A→C 23
gradient is worthy to notice: it indicates an increasing nutrient availability, allowing the rapid 24
growth and reproduction of plants which are unable to store nutrients in perennial organs 1
(Fédoroff et al., 2005).
2 3
The D group, which was mostly present in Lyons EF and Eawy EF (Figures, 3, 4), was 4
comprised of two species which already belonged to the A group: the fern A. filix-femina and 5
the perennial forb L. galeobdolon. Other species were the trees Acer pseusoplatanus and 6
Carpinus betulus, the shrubs Rubus idaeus, Ruscus aculeatus and Taxus baccata, the sedge- 7
grass Carex ovalis, the perennial forbs Anemone nemorosa, Callitriche sp., Digitalis 8
purpurea, Hyacinthoides non-scripta, Lotus pedunculatus, Polygonatum multiflorum, 9
Stellaria graminea, Stellaria nemorum and Viola riviniana, the annual forb Galeopsis tetrahit 10
agg., and the mosses Atrichum undulatum, Brachythecium rutabulum, Hypnum cupressiforme 11
agg., Isothecium myosuroides, Kindbergia praelonga and Plagiothecium denticulatum. Most 12
remarkable features are the diversity of mosses (see the projection of the moss richness vector 13
on Figure 3) and of spring-flowering geophytes. Several species (C. ovalis, Callitriche sp., L.
14
pedunculatus, Stellaria nemorum) are known for their affinity to wet environments. The D 15
group was better represented by the Endymio-Fagetum aretosum (Bardat, 1993), which is the 16
most hygrophilic sub-association of the Endymio-Fagetum, established in bottom woodland 17
within the study region. The D group of species was projected on the negative side of Factor 18
F3 (Fig. 3), which was positively correlated with weak disturbance and negatively correlated 19
with undisturbed ground cover, thereby indicating that corresponding plots were poorly rutted 20
by exploitation traffic. In the absence of straightforward data on management practices it may 21
be thought that wetter environments, given the difficulty of timber extraction with modern 22
vehicles, have been undisturbed for the last 50 years.
23 24
All species groups depicted by PCA were confirmed by RDA, on the basis of stand 1
properties (age, basal area, stem density, composition of canopy, subcanopy and shrub layers) 2
and ground disturbance (traffic cues, wood deposits). This seems to indicate that in the 3
context of the studied region ground vegetation responded to local factors which could be 4
described fairly well by a visual inspection of ground and above strata. As a corollary, this 5
points to the existence of a common regional species pool without any additional filter than 6
described by plot-scale local environment.
7 8
Once local environmental particularities have been ruled out by partialling out the 9
geographical position of sampling plots, a simpler pattern arose, with a single gradient of soil 10
fertility/acidity/light which was exemplified by Axis 1 of partial RDA (Fig. 6). The 11
distribution of species showed a gradient from acidophily to acido-intolerance which 12
corresponded to a continuum from the Ilici-Fagetum (species of the B group) to the Endymio- 13
Fagetum (species of A, C and D groups). The two dominant species of these habitats, 14
respectively I. aquifolium (Iaq) and H. non-scripta (Hno), were projected far from the origin, 15
on negative and positive sides of Axis 1, respectively. This gradient of decreasing acidity and 16
light and increasing fertility and Humus Index was associated with an increase in hornbeam 17
cover, both in canopy (Cov 1) and subcanopy (Cov 5). Whether hornbeam is a cause of 18
decreasing light for forest-floor vegetation is out of doubt, given its dense canopies strongly 19
select shade-tolerant species (Kwiatkowska et al., 1997). Whether hornbeam is a cause or a 20
consequence of variation in soil fertility/acidity and humus form does not deserve any 21
straightforward answer, although feed-back relationships can be suspected (Ponge, 2003).
22
Most studies comparing forest stands with and without hornbeam did not take into account 23
historical or environmental reasons why hornbeam was present or absent and they ascribed to 24
sylviculture only the observed effects (Aubert, et al. 2004; Decocq et al., 2005). Similar flaws 25
can be found in comparisons between coppice woods of varying nature and adjoining full- 1
grown forest stands (Hölscher et al., 2001). This lack of account for the spontaneous 2
establishment of C. betulus in natural or managed forests (Kwiatkowska et al., 1997;
3
Lawesson, 2000) is not so important for explaining variations in corticolous assemblages 4
(Bardat and Aubert, 2007) but it may flaw any conclusion about the impact of hornbeam on 5
forest-floor plant assemblages when mature forests of unknown past history are compared. As 6
shown by Decocq (2000) working on 157 sample plots distributed over the widest possible 7
range of geological substrates prevailing in northern France mixed-hardwood forests, forest- 8
floor vegetation was more influenced by the geological substrate than by the composition of 9
the overstory. However, when care is taken for substrate conditions being identical, the 10
positive impact of hornbeam on the soil fertility of beech forests, which is mainly due to 11
better litter quality of hornbeam compared to beech (Lemée and Bichaut, 1973), and 12
consequent influences on forest-floor vegetation, can be assessed less ambiguously (Decocq 13
et al., 2004; Ponge and Chevalier, 2006; Van Calster et al., 2007). Awaiting further 14
clarification of the ‘hornbeam’ effect, it can be hypothesized that the combined influence of 15
geological substrate, natural establishment of hornbeam and sylviculture (in particular 16
coppicing) results in a feed-back loop in favour of mull humus forms and associated 17
vegetation.
18 19
The variation of ground vegetation can thus be explained by a combination of humus 20
quality, soil fertility, soil acidity and light along a single environmental gradient from poorly 21
productive (and species-poor) to highly productive (and species-rich) forest ecosystems, in 22
accordance with some (but far from all) theoretical studies (Ulrich, 1994; Bruno et al., 2002) 23
and in line with experimental and descriptive studies (Falkengren-Grerup and Tyler, 1993;
24
Brunet et al., 1997; Rodríguez-Loinaz et al., 2008). To these rather simple effects, due to 25
multiple but strongly interconnected factors stemming in a limited number of stable states 1
(Perry, 1995; Ponge, 2003; Beisner et al., 2003), are superimposed local effects of a more 2
complex (i.e. harder to discern) nature. Climate (Bakkestuen et al., 2008), atmospheric 3
deposition (Brandt and Rhoades, 1972; Falkengren-Grerup, 1986), past history (Peterken and 4
Game, 1984; Koerner et al., 1997; Gachet et al., 2007), management practices (Thomas et al., 5
1999; Decocq, 2000; Godefroid and Koedam, 2004), dispersal from sources of migration 6
(Björkman and Bradshaw, 1996; Bossuyt et al., 1999) and succession (Watt, 1934; Myster 7
and Pickett, 1992; Godefroid et al., 2005) are supposed to act more or less independently in 8
different forests (McCune and Allen, 1985), thereby creating the variegated forest habitats 9
which could be described (but not necessarily fully explained) locally (Rol, 1937; Ricklefs, 10
1987; Huston, 1999).
11 12
Acknowledgements
13 14
The study was financially supported by a grant given to the junior author by the Office 15
National des Forêts (ONF). Authors are grateful to local authorities for access to the sites and 16
commodities and to Thierry Gautrot his help during field work. Luc Croisé (ONF) and 17
Élisabeth Duguin (Météo-France) are acknowledged for providing data about atmospheric 18
deposition and climate, respectively.
19 20
References
21 22
Aubert, M., Bureau, F., Alard, D., Bardat, J., 2004. Effect of tree mixture on the humic 23
epipedon and vegetation diversity in managed beech forests (Normandy, France). Can.
24
J. For. Res. 34, 233-248.
25
1
Aubin, I., Gachet, S., Messier, C., Bouchard, A., 2007. How resilient are northern hardwood 2
forests to human disturbance? An evaluation using a plant functional group approach.
3
Écoscience 14, 259-271.
4 5
Austin, A.T., Yahdjian, L., Stark, J.M., Belnap, J., Porporato, A., Norton, A., Ravetta, D.A., 6
Schaeffer, S.M., 2004. Water pulses and biogeochemical cycles in arid and semiarid 7
ecosystems. Oecologia 141, 221-235.
8 9
Bakkestuen, V., Erikstad, L., Halvorsen, R., 2008. Step-less models for regional 10
environmental variation in Norway. J. Biogeogr. 35, 1906-1922.
11 12
Bardat, J., 1993. Phytosociologie et écologie des forêts de Haute-Normandie. Leur place dans 13
le contexte sylvatique ouest-européen. Bull. Soc. Bot. Centre-Ouest NS 11, 1-376.
14 15
Bardat, J., Aubert, M., 2007. Impact of forest management on the diversity of corticolous 16
bryophyte assemblages in temperate forests. Biol. Conserv. 139, 47-66.
17 18
Bartoli, M., Tran-Ha, M., Largier, G., Dumé, G., Larrieu, L., 2000. ECOFLORE, un logiciel 19
simple de diagnostic écologique. Rev. For. Fr. 52, 530-546.
20 21
Beisner, B.E., Haydon, D.T., Cuddington, K., 2003. Alternative stable states in ecology.
22
Front. Ecol. Environ. 1, 376-382.
23 24
Belya, L.R., Lancaster, J., 1999. Assembly rules within a contingent ecology. Oikos 86, 402- 1
416.
2 3
Bergès, L., Gégout, J.C., Franc, A., 2006 Can understory vegetation accurately predict site 4
index? A comparative study using floristic and abiotic indices in sessile oak (Quercus 5
petraea Liebl.) stands in northern France. Ann. For. Sci. 63, 31-42.
6 7
Björkman, L., Bradshaw, R., 1996. The immigration of Fagus sylvatica L. and Picea abies 8
(L.) Karst. into a natural forest stand in southern Sweden during the last 2000 years. J.
9
Biogeogr. 23, 235-244.
10 11
Bossuyt, B., Hermy, M., Deckers, J., 1999. Migration of herbaceous plant species across 12
ancient-recent forest ecotones in Central Belgium. J. Ecol. 87, 628-638.
13 14
Brandt, C.J., Rhoades, R.W., 1972. Effects of limestone dust accumulation on composition of 15
a forest community. Environ. Pollut. 3, 217-225.
16 17
Britton, A., Marrs, R., Pakeman, R., Carey, P., 2003. The influence of soil-type, drought and 18
nitrogen addition on interactions between Calluna vulgaris and Deschampsia flexuosa:
19
implications for heathland regeneration. Plant. Ecol. 166, 93-105.
20 21
Brunet, J., Falkengren-Grerup, U., Tyler, G., 1997. Pattern and dynamics of the ground 22
vegetation in south Swedish Carpinus betulus forests: importance of soil chemistry 23
and management. Ecography 20, 513-520.
24 25